Currently, users find it difficult to find and install facility software on their own computers, even when they have compatible operating systems. There is a trend towards people installing software by themselves and not having the use of computer professionals, which parallels the trend towards using PCs running Windows and Linux. If facility programs are provided on FTP sites, ideally binaries should be available, and if not, the compilation process should be simple and involve a minimum of libraries. For example, some GNU freeware can be difficult to install on non-Linux machines due to the number of system tools that need to be installed first. Also, the Hierarchical Data Format (HDF) libraries are daunting to install on VMS.
Presumably, the new data format will be relatively modern in design, using hierarchical data formats concepts, and be portable to all computer platforms so that it will have a long lifetime. To make it easy to read in the new data format into user's programs, an application program interface (API) should be available in Fortran and C. This would consist of a series of library calls to get the title, number of histograms, contents of the histograms, etc. Ideally, this API would be simple to install and use.
The third issue is the availability of browser or math programs which can read in our data format. If we use one which is unique, such as MUD, we are on our own, except to the extent we can divert facility programming resources to extend programs such as Physica and OpenGenie to read it in. If we use a standard HDF data format (like NeXus), then we can use pre-existing neutron scattering facility software such as OpenGenie and commercial software such as the SciSpy Browser and the data visualization program IDL. (NeXus is a data format for neutron and x-ray scattering written in HDF which is eventually intended to become their universal standard.) The benefit of these programs is that they handle a wide range of terminal types, have good graphics, exist for a variety of operating systems (but not all), and, in the case of programs like IDL, one can write macros to manipulate the data. Macros should be more efficient to ``program'' than writing code from scratch in Fortran or C. Plus, someone else is maintaining the main program. Choosing to write our data format in HDF would give us these benefits automatically.
However, there are some practical drawbacks to HDF. First of all, people are not generally already using those programs like IDL and OpenGenie which already have HDF capabilities (see Table 8). People will probably initially resent any pressure to use an additional, expensive program. While OpenGenie is free for academic use, IDL is on the expensive end of the price range for math programs. Matlab/Octave also has HDF capabilities, but I am not knowledgeable about other programs. We need to make sure that we strike a good balance between the desire to keep things quick and easy for the users and the goal of selecting a modern data format which will have a long life span.